← Back to home
Agentic Projects Archive — Building with AI Things I've built by vibe-coding with AI agents — a World Cup lineup builder, a daily player-guessing game, a dog-walking app, an interview simulator, a Hollywood agent game, and a model-comparison tool. Select any project to read the full story.
Starting XI 2026 — lineup builder SubAgents Database Design Cursor Claude Code
Starting XI 2026 A crowdsourced lineup builder for the 2026 World Cup — fans pick a formation and starting XI from real player data, and the site tallies the crowd's choices by country. Idea to public launch in five days.
2026 Readership Retention Experiments — summarizer + feed Eval Harness Cursor Anthropic API
Readership Retention Experiments An audit of an existing sports journalism app built around one question: how would I improve retention? I shipped an AI summarizer with five distinct reading modes and a prototype feed built to pull readers back at their peak moments.
2026 World Cup Player Picker — daily puzzle Analytics Game Sports
World Cup Player Picker A daily guessing game built on the cleaned player data from Starting XI 2026 — one puzzle a day to test how well fans really know the tournament's squads. Idea to live in about three hours.
2026 SafePaws — walk recommendations MCP iOS Cursor Claude Code
SafePaws A native iOS app that turns live weather and air-quality data into safe dog-walking windows — tailored by breed, age, weight, and health conditions.
2026 — NOW Interview Simulator — XP desktop LLM Scoring Lovable Claude Code
PM Interview Simulator A voice-enabled PM interview simulator dressed as Windows XP — practice out loud, get each answer scored 0–100, and track your progress.
2026 Hollywood Agent — deal in progress Prompt Engineering Anthropic API
Hollywood Agent Simulator A text-based game where you play a Hollywood super-agent — negotiate with studios and talent through dynamic, LLM-driven characters à la Ari Gold.
2026 LLM Tester — side-by-side compare LLM Evaluation Claude API OpenAI API Gemini API
LLM Tester A side-by-side LLM comparison tool built for PMs — run one prompt across providers and models, rate the outputs, and let Claude summarize which model fits which use case.
2026